Team05: Analysis of UP politicians twitter activity before and after UP elections 2022
Computational Social Science
Analysis of UP politicians twitter activity before and after UP elections 2022.
Problem Statement
In today's high tech world, political parties take help from social media platforms like twitter to increase the number of votes which they will get in elections.
We are analyzing the activities of political leaders of different parties and how these activities are affecting the results of elections and the perspective of people.
Why is it related to CSSS
Computational social science is the study of social phenomena using digitized information and computational methods.
Social Phenomena:
A political party and its leaders are representatives of the society and since, and bringing them into power directly impacts the society.
Digitized Information:
We have collected tweets of politicians before and after the 2 months duration of the elections.
Computational methods:
We are using statistical methods to find out the impact of the political leaders in elections.
Attention gained by the BJP through twitter activity of its UP politicians
Methodology pipeline:
To address this objective we examined the twitter activity of the BJP politicians. We have selected ten politicians based on the followers count in twitter. They are the top ten most followed BJP politicians on twitter. These politicians are those who contested in UP elections 2022.
We collected all the tweets before the election in the period of december 2021 to march 2022. We manually went through all the tweets and identified the tweets which are related to elections and we labeled the tweets according to the category whether they are in favor of their party or against some other party.
Figure1: Count of followers of the BJP UP politicians
Mean likes gained by the tweets
From this analysis we got to know who are the people gaining attention in the form of likes.
Experimental Design and results
We did the analysis on the tweets favoring their own party and criticizing other parties before and after elections.
For all the below graphs the x-axis represents the count of the total tweets collected and y-axis represents the count of tweets which are related to the election
TWEETS FAVORING THE PARTY BEFORE ELECTIONS
The tweet posted falls in the category of favoring their own party before the elections. And the size of the bubble represents the mean like gained on the tweets of the politicians.
We observed the politicians with a high number of followers like Yogi Adityanath, Smriti Irani, KP Maurya are able to gain more average likes, along with them Aditi Singh is having more average likes on the tweets even with less number of followers compared to Smriti Irani, KP Maurya .
Figure2: Mean likes gained by the tweets of BJP UP politicians which favors their party before elections
TWEETS CRITICIZING OTHER PARTIES BEFORE ELECTIONS
The tweets posted are against the other parties before the elections. And the size of the bubble represents the mean like gained on the tweets of the politicians.
From the graph we observed that the top most followers politicians like Yogi Adityanath, Smriti Irani, KP maurya and PT Shrikant are able to gain more average of likes on the tweets which are against the other parties as compared to the other politicians.
Figure3: Mean likes gained by the tweets which are against other parties
by BJP UP politicians after elections
TWEETS FAVORING THE PARTY AFTER ELECTIONS
The tweet posted falls in the category of favoring their own party after the elections. And the size of the bubble represents the mean like gained on the tweets of the politicians.
We observed that only the top most followed leaders are able to gain more likes on average. Compared to before the elections, AditiSingh mean likes gain decreased whereas remaining politicians maintained their positions.
Figure4: Mean likes gained by the tweets of BJP UP politicians which favors their party after elections
What we tried and what didn’t work out
We tried to identify the tweets after the elections and are against other parties. But most of the tweets after the election are related to their own party. So we excluded that one from the analysis. Also we tried to do the same analysis on mean retweet count. The results are almost the same as that of the mean likes count. That’s why we didn’t include them.
Deductions / Discussions
From the above analysis that we got to know that along with the top most followed politician Yogi Adityanath other least followed people compared to Yogi such as Aditi Singh, PT Shrikant are able to tweet in such a way that those tweets related to favoring the party attract more people and gain more likes. Whereas when it comes to tweets which are against the other parties only the top most followed people are able to make those tweets and gain likes. After the elections except Aditi Singh, KP Maurya remaining everyone maintained their position in the mean like gain.
Tweet frequency of the BJP Politicians
Experimental Design and results
We did the analysis on the tweets before and after the election on the tweets which are related to elections.
BEFORE ELECTIONS
In the below graph the x-axis represents the dates which are before the election and the y-axis represents the count of tweets. Each line in the graph is corresponding to the top BJP politicians.
The graph represents the frequency count of tweets by the top BJP politicians on each date before the elections. In this analysis we can understand which politician is engaged more in twitter during the before election period.
From the graph we can see that Yogi Adityanath, KP Maurya, Brajesh Pratap and Vinood Sonkar have more frequency of tweets per day before the election period.
Figure5: Frequency count of tweets before the election
AFTER ELECTIONS
In the below graph the x-axis represents the dates which are after the election and the y-axis represents the count of tweets. Each line in the graph is corresponding to the top BJP politicians.
The graph represents the frequency count of tweets by the top BJP politicians on each date after the elections. In this analysis we can understand which politician was engaged in posting tweets on twitter after the election.
From the graph we can observe that KP Maurya and Vinood Sonkar have more frequency of tweets per day after the election period.
Figure6: Frequency count of tweets after the election.
From both the analysis we can see that the frequency of election related tweets of Yogi Adityanath and Brajesh Pratap have reduced after the elections.
Words mostly used by politicians in their tweets.
We selected a few political leaders and created a word cloud of the most frequent words used by these political leaders in their tweets and we analyzed how many words are related to elections and how they are favoring their own party or opposing any other party. We find these assumptions on the tweets before and after the elections.
BEFORE ELECTIONS
Myogiadityanath
AditiSinghRBL
Kpmaurya1
Smritiirani
AFTER ELECTIONS
Myogiadityanath
AditiSinghRBL
Kpmaurya1
smritiirani
From the above graphs of common words of individual political leaders, the main keywords before the elections which are related are as follows - विकास, सर्कार, भाजपा, वोट, निर्माण, योजना, #upफिरमंगेभजपा, #upमंIगेभजपा, योगदान, आशीर्वाद, विश्वास, जनता, कांग्रेस
The main keywords after elections which are related are as follows - धन्यवाद्, शुभकामनाये, स्वागत, उज्वल, बधाई, शपथ, सफलता, आगमन, परिणाम
We can infer from this that the words which are used in the tweets before election are more inclined to motivate people to vote for BJP and promising type words which promises to work for the people of UP, development of UP, showing trust to the party, telling the audience to vote for BJP and again want BJP government in UP, focuses on making policies for farmers, work on unemployment and poverty problems, work for pension of old peoples,security of girls. Whereas the words which are tweeted after election results are more related to congratulating BJP for winning the election like congratulations, happiness, results, succession elections, welcome again, village, gifts, people, bright future. Some political leaders like Aditi Singh and Smriti Irani posted less content on Twitter which is related to elections which is the reason we are getting very less unique words in the graph.
Political parties and their different target audience
Methodology pipeline:
To address this objective we examined the tweets of the 11 BJP and 9 SP politicians. Then we manually curated a list of words in hindi for 8 categories of target audiences. They were Women, Farmers, groups based on caste (Brahmins,Hindus , Muslims), Backward groups like(Dalits, OBC, EWS) , Students, Youth , Poor, Old and disabled. We collected the tweets containing these words to analyze if a party targets some specific group of people. To do so we first did the preprocessing of the tweets which included removal of emojis, generation of tokens, removal of ‘\n’, removal of stopwords and punctuations, removal of frequent unnecessary hindi words.
Experimental Design and results
We did the analysis for before and after elections.
BEFORE ELECTIONS
The graph is as follows:-
AFTER ELECTIONS
The graph is as follows:
The x axis contains the names of different target groups which have been analyzed.The y axis has the tweet ratios .The tweet ratio denotes that the number of tweets for any certain group have been divided by the total number of tweets done by the party. It is done so as to nullify the effect of higher tweet count of BJP as compared to SP .It is primarily due to two major reasons. Firstly, there are 11 BJP and 9 SP politicians taken into consideration. Secondly, the tweet count of BJP members is higher than the SP members.
Results:-
The major results that can be drawn from the above two graphs are as follows:-
For both the scenarios, before and after elections the major tweet ratio for Samajwadi Party is dedicated to youth
Before elections BJP targeted the most on the Poor, and the second most after Youth in the period after elections.
For four classes, (Women, Farmers, Poor and Old and disabled) , the results inverted for both the parties.For eg, It means that before elections, the tweet ratio for BJP was higher for ( women,Farmers,Poor ) and after elections , it became higher for SP. For Old and disabled, it was higher for SP before elections and after elections it became higher for BJP.
It can be observed that both the parties focussed the least on the Old age group.
What we tried initially and what didn’t work out
We tried including the rally speeches of both the parties by trying to convert them into text ,but it was not sufficient to gather those many speeches which could cover all the target groups.
Deductions / Discussions
According to the first result, it can be denoted that mostly both the parties focussed mainly on targeting the youth audience as they might want to target the largest part of India’s population (22 %) around more than 261 million people. The inversion of results for some target groups indicate that the parties paid less emphasis on these groups post elections.The old age group has been paid the least attention by both the parties, which denotes that they constitute the least share of discussions amongst the hot party topics.
Election results revolve around Top leaders of different political parties
Methodology pipeline:
To solidify the claim of our hypothesis, we did a survey. We have listed major leaders contesting in the election from Uttar Pradesh and some leaders that are well known in the party but not so popular around the people. To collect these leader's lists we have used parameters like the number of Twitter followers. Then we shared the form among IIIT students. To our surprise, we have figured out that most people know top leaders. And people vote mainly for the people they know or can connect to some people in the party. For example PM Narendra Modi with BJP, Rahul Gandhi with INC, etc. So, people will vote for the party they can connect with. Since most of them knew only top Leaders, top leaders played a major role in deciding the result of the election. Survey form Link. Survey Results Link.
Experimental Design
We have used the survey to understand the familiarity of the people with the politicians in UP. We have also done the analysis for before and after elections.
The demographic from the response has some variance but there is a good representation of the people from UP. We intentionally asked people from UP to fill out the form, so the responses are high from UP.
To measure the impact of political leaders we have taken four parameters (with before and after election values):
Number of Tweets: We have collected all the tweets from the duration of “1st Jan 2022 to 10th March 2022” (Before Election) and “11th March to 6th April” (After Election). The number of tweets measures how the politician tries to gather attention on Twitter but it doesn’t tell us anything about the impact on the users on Twitter.
Number of Mention: People will mention people mostly if they know them/seeking help in some matter. Any user will mention only people whom they know. So we have taken a few members from BJP and one from SP and INC. And found out that mostly the top leaders have the highest mention, before and after the election. Although after the election activity of the users decreased by a lot, relatively top leaders have still higher mentions than others.
Reply Count: There is a difference between Reply Count and Retweet Count. The reply is the response to something written by someone on Twitter. The retweet is the way you forward another user's tweet to your followers. So Reply count can be treated as a measure of the bond between the politician and his/her Twitter followers.
Tweet Quote Count: A quote tweet is a retweet with an added comment that allows you to add your own spin on the retweet while still giving the original post-exposure. It's a great way to share other people's content while supporting them and putting a little twist on it that's all your own.
What we tried and what didn’t work out
I tried to create a classifier based on tweets that are from the top leader and local leaders. My idea was to focus on the tweet content by the leaders. But mostly all the people from the same party tend to repeat the same agenda. So the results are not accurate for the classifier.
We tried to share the form outside the IIIT, but we need to explain to people a lot of things, and lots of time is getting wasted. So we restricted ourselves to IIIT students only.
Inferences
Most of the people only knew the top leaders which are evident from the Google Form response that we have collected over a diverse set of Group. So voting for a person whom we don’t know is very unlikely.
The “Tweet Count” of top leaders is high. Along with that the percentage of “Mention Count” is very high which shows the responsiveness of users to the Top Politician Leader.
Top Leaders are getting more “Reply Count” than other leaders. The percentage of "Tweet Quotes" of top leaders is high. This signifies that people are giving more importance to the Top Leader of the political party, and they are the main face of the party in the broader sense.
Hence, We can conclude that Elections are in some sense decided by top political leaders.
Future Work:
We can scale the number of responses to more people.
For the reply and quote count we can try to add all the leaders active on Twitter rather than shortlisting with some parameters and working with them.
Comments
Post a Comment